Logo ROOT   6.16/01
Reference Guide
TActivationTanh.cxx
Go to the documentation of this file.
1// @(#)root/tmva $Id$
2// Author: Matt Jachowski
3
4/**********************************************************************************
5 * Project: TMVA - a Root-integrated toolkit for multivariate data analysis *
6 * Package: TMVA *
7 * Class : TActivationTanh *
8 * Web : http://tmva.sourceforge.net *
9 * *
10 * Description: *
11 * Tanh activation function (sigmoid normalized in [-1,1] for an ANN. *
12 * *
13 * Authors (alphabetical): *
14 * Matt Jachowski <jachowski@stanford.edu> - Stanford University, USA *
15 * *
16 * Copyright (c) 2005: *
17 * CERN, Switzerland *
18 * *
19 * Redistribution and use in source and binary forms, with or without *
20 * modification, are permitted according to the terms listed in LICENSE *
21 * (http://tmva.sourceforge.net/LICENSE) *
22 **********************************************************************************/
23
24/*! \class TMVA::TActivationTanh
25\ingroup TMVA
26Tanh activation function for ANN.
27*/
28
30
31#include "TMVA/TActivation.h"
32
33#include "TMath.h"
34#include "TString.h"
35
36#include <iostream>
37
39
40////////////////////////////////////////////////////////////////////////////////
41/// a fast tanh approximation
42
44 if (arg > 4.97) return 1;
45 if (arg < -4.97) return -1;
46 float arg2 = arg * arg;
47 float a = arg * (135135.0f + arg2 * (17325.0f + arg2 * (378.0f + arg2)));
48 float b = 135135.0f + arg2 * (62370.0f + arg2 * (3150.0f + arg2 * 28.0f));
49 return a/b;
50}
51
52////////////////////////////////////////////////////////////////////////////////
53/// evaluate the tanh
54
56{
57 return fFAST ? fast_tanh(arg) : TMath::TanH(arg);
58}
59
60////////////////////////////////////////////////////////////////////////////////
61/// evaluate the derivative
62
64{
65 Double_t tmp=Eval(arg);
66 return ( 1-tmp*tmp);
67}
68
69////////////////////////////////////////////////////////////////////////////////
70/// get expressions for the tanh and its derivative
71/// whatever that may be good for ...
72
74{
75 TString expr = "tanh(x)\t\t (1-tanh()^2)";
76 return expr;
77}
78
79////////////////////////////////////////////////////////////////////////////////
80/// writes the Tanh sigmoid activation function source code
81
82void TMVA::TActivationTanh::MakeFunction( std::ostream& fout, const TString& fncName )
83{
84 if (fFAST) {
85 fout << "double " << fncName << "(double x) const {" << std::endl;
86 fout << " // fast hyperbolic tan approximation" << std::endl;
87 fout << " if (x > 4.97) return 1;" << std::endl;
88 fout << " if (x < -4.97) return -1;" << std::endl;
89 fout << " float x2 = x * x;" << std::endl;
90 fout << " float a = x * (135135.0f + x2 * (17325.0f + x2 * (378.0f + x2)));" << std::endl;
91 fout << " float b = 135135.0f + x2 * (62370.0f + x2 * (3150.0f + x2 * 28.0f));" << std::endl;
92 fout << " return a / b;" << std::endl;
93 fout << "}" << std::endl;
94 } else {
95 fout << "double " << fncName << "(double x) const {" << std::endl;
96 fout << " // hyperbolic tan" << std::endl;
97 fout << " return tanh(x);" << std::endl;
98 fout << "}" << std::endl;
99 }
100}
#define b(i)
Definition: RSha256.hxx:100
double Double_t
Definition: RtypesCore.h:55
#define ClassImp(name)
Definition: Rtypes.h:363
Tanh activation function for ANN.
TString GetExpression()
get expressions for the tanh and its derivative whatever that may be good for ...
Double_t Eval(Double_t arg)
evaluate the tanh
Double_t fast_tanh(Double_t arg)
a fast tanh approximation
Double_t EvalDerivative(Double_t arg)
evaluate the derivative
virtual void MakeFunction(std::ostream &fout, const TString &fncName)
writes the Tanh sigmoid activation function source code
Basic string class.
Definition: TString.h:131
Double_t TanH(Double_t)
Definition: TMath.h:645
auto * a
Definition: textangle.C:12